Large language models produce semantically rich embeddings, yet direct employment of generic LLM embeddings fails to satisfy news recommendation demands due to inherent semantic mismatches with task targets. Existing fine-tuning methods including LoRA can narrow such gaps but bring prohibitive computational overhead, restricting real-world deployment. This work proposes lightweight task-adaptive modules (TAMs). It keeps LLM parameters fixed and transforms offline embeddings into task-specialized representations without full-model backpropagation, drastically cutting training costs. Evaluated on MIND benchmarks across ten mainstream recommendation architectures, TAMs achieves comparable accuracy to LoRA, with computational cost reduced to 1/10 of LoRA’s level. TAMs outperform GloVe-based models by 1.3–12.4% in AUC and scale effectively to MINDlarge. Ablation experiments confirm that the nonlinear projection is pivotal to performance improvement, and statistical validation across three random seeds confirms result robustness. This paradigm provides an efficient low-cost solution for LLM-based news recommendation under resource constraints.

With the emergence of online public opinion leading to the public opinion war, a new ideological field struggle, monitoring online public opinion is an important measure to maintain long-term social stability. This paper studied the social network modeling and Multiagent technology, analyzed the Internet public opinion diffusion and communication model, used the programmable modeling environment Netlogo combined with Multiagent technology to simulate the real public opinion communication process, designs a Multiagent-based public opinion communication model, analyzes and verifies the effectiveness of the model by experiments.

With the emergence of massive news every day leading to information overload, it is difficult for people to select the content they are really interested in from numerous news articles. The language knowledge and strong generalization ability of the large model itself provide the possibility of intelligent news recommendation. This paper does not fine-tune the big language model, but uses the prompt engineering technology of the big language model to design a hybrid prompt engineering instruction combining prompt words, context learning and thought chain prompt to stimulate the recommendation performance of the big language model, and realizes the intelligent news recommendation of the big language model.